An analysis published in September 2026 in the journal PLOS Medicine found associations between sleep characteristics and the future occurrence of diseases in a cohort of 95,559 UK Biobank participants. More estimated time in REM sleep was associated with a lower observed risk of 83 conditions. The study does not demonstrate that increasing REM reduces the risk of a disease in a specific person.
What the team analyzed
Participants wore wrist accelerometers for seven nights. Using a deep learning algorithm called SleepNet, the researchers estimated sleep stages, total duration, night-to-night irregularity, and time awake after sleep onset. They then cross-referenced these measures with health records to assess 1,049 outcomes over a median follow-up of 8.9 years.
After statistical adjustments, the study identified 156 significant associations between sleep patterns and incident diseases. A difference of approximately 48 minutes in the REM measure used by the researchers was associated with a lower risk of 83 conditions. More deep sleep was associated with a lower risk of seven conditions. Sleep duration also had nonlinear relationships with several outcomes, with estimates of lower risk often concentrated between six and eight hours.
Association is not causation
The authors themselves emphasize that the observational design does not allow the conclusion that differences in sleep caused the differences in risk. Unmeasured factors, diseases not yet diagnosed, and health conditions that alter sleep may influence the results. Measurements were collected over just one week, and sleep stages were estimated using an accelerometer and an algorithm, not polysomnography.
The UK Biobank sample also does not perfectly represent the entire population: the researchers point to volunteer selection characteristics and limitations in generalizability. Therefore, the numbers describe associations within the cohort studied; they are neither an individual prediction nor a REM sleep target to pursue on a smartwatch.
How to use this information
Consumer devices can estimate movements and nighttime patterns, but they do not confirm a sleep disorder or replace testing recommended by a healthcare professional. If there is persistent difficulty sleeping, significant sleepiness, snoring, or observed pauses in breathing, it is worth seeking a clinical assessment rather than trying to interpret an app’s graphs in isolation.
Scientific source
- Luo et al. Accelerometer-derived real-world sleep stages and risk of incident diseases. PLOS Medicine, September 17, 2026.
- Research announcement distributed by EurekAlert! on behalf of PLOS.
Journalistic and educational content. The findings do not diagnose diseases and do not replace medical evaluation.